Nvidia insists it isn’t Enron, but its AI deals are testing investor faith | Nvidia

Nvidia is, in important ways, nothing like Enron, the Houston energy giant that collapsed in 2001 amid multibillion-dollar accounting fraud. It’s also not like companies like Lucent or Worldcom, which went bankrupt during the dotcom bubble.

But the fact that he has to reiterate this to his investors is less than ideal.

Nvidia, now worth more than $4 trillion (£3 trillion), produces the proprietary technology powering the world’s rise in artificial intelligence: software packages and silicon chips that train and host systems such as ChatGPT. Its products fill data centers from Norway to New Jersey.

This has been a phenomenal year for the company: At least $125 billion in deals have been made to ease its access to the PC market, from a $5 billion investment in Intel to a $100 billion investment in OpenAI, the startup behind ChatGPT.

But even as those deals fueled rising stock prices and set the stage for CEO Jensen Huang’s energetic world tour, doubts have emerged about how Nvidia does business, especially as it becomes increasingly central to the health of the global economy.

The beginning of these concerns was the cyclical nature of many deals. These arrangements are similar to vendor financing: Nvidia lends money to customers so they can buy its products.

The biggest of these is its deal with OpenAI, which involves Nvidia investing $10 billion in the company every year for the next 10 years; Most of this amount will go towards purchasing Nvidia’s chips. Another is its deal with CoreWeave, which provides on-demand computing capacity to major AI firms, essentially leasing Nvidia’s chips.

The cyclicality of these deals has led to comparisons with Lucent Technologies, a telecom company that aggressively lent money to its customers but overextended itself and unraveled in the early 2000s. Nvidia has aggressively refuted any claims of similarity, saying in a recently leaked memo that it “does not rely on vendor financing arrangements to drive revenue.”

James Anderson, a well-known tech investor, describes himself as a “big fan” of Nvidia but said this year that the OpenAI deal offers “more reasons than ever before to be worried about it.”

He added: “I have to say that the words ‘supplier financing’ don’t carry very pleasant resonances for someone my age. It doesn’t sound like what many telecoms suppliers were doing in 1999-2000, but it does have some rhyme. I don’t think it makes me feel entirely comfortable in that respect.”

Other recent high-profile deals include tech firm Oracle spending $300 billion on OpenAI data centers in the US, and ChatGPT developer paying back almost the same amount for using those data centers. In October, OpenAI and chipmaker AMD signed a multibillion-dollar chip deal that also gave OpenAI the option to buy a stake in rival Nvidia.

There is also a deal with CoreWeave in which OpenAI received $350 million in CoreWeave shares, in addition to a commitment to buy $22 billion of data center capacity from the cloud provider. Asked this month about cyclicality in the AI ​​industry, CoreWeave CEO Michael Intrator said: “Companies are trying to address the drastic change in supply and demand. You can do this by working together.”

All these moves form part of OpenAI’s $1.4 trillion bet on computing capacity to build and operate models that it claims will transform economies and bring back that spending. OpenAI argues that although there is an investment component to the Nvidia and AMD deals, it comes into play only after chips are purchased and distributed, while investments create aligned incentives to build AI infrastructure at scale.

Nvidia also used builds called special purpose vehicles (SPVs) in financing agreements. The best-known example is the SPV linked to Elon Musk’s xAI: an entity in which Nvidia has invested $2 billion, the money will be used to buy Nvidia’s chips.

This has led to comparisons with Enron, which used SPVs to keep debt and toxic assets off the balance sheet, convincing investors and creditors that it was stable while hiding ballooning liabilities.

Nvidia has also vehemently denied that it resembles Enron: In the same leaked memo discussing Lucent, it said its reporting was “complete and transparent” and that “unlike Enron,” it “does not use special purpose assets to hide debt and inflate revenue.”

Journalist Ed Zitron, who is skeptical of the artificial intelligence boom, also says that Nvidia agrees on this issue. It is unlike either company. Unlike Lucent, he says, it doesn’t appear to be taking on a lot of debt to finance its cyclical deals, and many of the clients it backs aren’t as risky as Lucent’s dotcom bubble partners. Zitron argues that it is not like Enron because it is highly transparent about its complex, off-balance sheet dealings.

So what might warrant a comparison? Nvidia “isn’t hiding its debt, but it’s leaning heavily on vendor-funded demand, which is at risk if AI growth slows,” says Charlie Dai, an analyst at research firm Forrester. “The concern is sustainability, not legality.”

Fundamentally, whether Nvidia can continue its descent depends on whether AI actually takes off, generating billions for its enterprise users, and whether companies like OpenAI, Anthropic, and CoreWeave (Nvidia’s customers) can go completely aground and continue buying up their systems. This possibility alone is debatable. If that doesn’t happen, Dai says, Nvidia “could face devaluation of equity shares and outstanding receivables”: meaning it could lose a lot of money and its stock price could fall.

Reached for comment, an Nvidia spokesperson told the Guardian that chief financial officer Colette Kress had made a statement to investors in early December. Kress said they don’t see an AI bubble, instead pointing to trillions of dollars of business ahead of Nvidia over the next decade.

Kress specifically argued that Nvidia’s recent – huge – deals are just the beginning for the company, and that the real money will be made in the coming years, largely by replacing nearly all chips in existing data centers with its own products.

There’s another complication, which is that the health of Nvidia, and thus the health of the entire global economy, also depends on whether AI launches in time for Nvidia and its customers to pay off debt from massive data center deployments and significant capital expenditures.

Add to this a final category of concern: recent big-ticket deals worth billions of dollars with vague terms with countries like South Korea and Saudi Arabia. In October, Nvidia announced that it would supply 260,000 Blackwell chips to the South Korean government and South Korean companies. the value of this deal It has not been disclosed but is estimated to be in the billions.

Likewise with Saudi Arabia. State-owned artificial intelligence startup Humain has committed to distribute up to 600,000 Nvidia chips: again, it was not announced when this distribution would include actual purchases and at what price. Nvidia has a number of other such strategic partnerships, including Italy, French AI champion Mistral and Deutsche Telekom; all of which contain thousands of chips and unknown sums.

Governments are likely to pay. There is nothing circular about a sovereign partnership with Germany. But the deals imply considerable uncertainty, nestled in a difficult web of commitments that require large capital expenditures and are based on ambitious assumptions that the economy will undergo a revolution in the coming years.

“They concentrate the risk on a few large customers,” says Dai. “If implementation delays occur, Nvidia’s revenue recognition and cash flow may be impacted.”